The use of early, aggressive broad-spectrum antibiotic therapy is common in hospitalized patients. Antibiotic de-escalation broadly refers to the discontinuation of antibiotics that are providing a spectrum of activity greater than necessary, or switching to narrower spectrum therapy once a patient is stable. Despite endorsement of antibiotic de-escalation by practice guidelines, limited data characterize the application of antibiotic de-escalation in practice. Furthermore, researchers have subjectively measured antibiotic de-escalation, which has prevented comparison of de-escalation rates between hospitals, and hindered evaluation of factors that may explain inter-hospital variability in de-escalation rates. Proposal aims include: 1) development and validation of a metric that characterizes the relative bacterial spectrum of antibiotics which can be used to measure antibiotic de-escalation; 2) application of the de- escalation metric to a cohort of veterans with health care associated pneumonia (HCAP) allowing comparison of antibiotic de-escalation rates between hospitals; and 3) identification of factors that explain inter-hospital variability in antibiotic de-escalation rates. In aim 1 a Delphi process utilizing a national pool of pharmacy and physician experts who practice antibiotic de- escalation will: a) develop a spectrum scoring metric to measure spectrum of antibiotic activity; and b) define antibiotic de-escalation according to criteria based on the scoring metric. The Delphi process output will be applied to develop a rule to classify antibiotic de-escalation status in electronic medical records-based antibiotic administration data. Internal validity will be assessed and calibration of the de-escalation rule to optimize the prediction of expert opinion will be accomplished by comparing de-escalation judgments of clinicians to rule based de- escalation. In aim 2, the antibiotic de-escalation rule will be applied to a retrospective cohort of patients admitted with health care associated pneumonia (HCAP) in 124 VAMC inpatient facilities nationwide. Antibiotic de-escalation rates will be expressed at the hospital level, stratified based on facility characteristics such as size and complexity level. In aim 3, a mixed- effects multivariable logistic model of antibiotic de-escalation probability will be developed to better understand the variability in de-escalation rates between hospitals. Adjusting for covariates selected in the validated model will allow associations between antibiotic de- escalation and bacterial culturing rates to be evaluated. The proposed research has significant potential to improve health by providing a basis for understanding relationships between bacterial spectrum, culturing practices, and treatment decisions such as antibiotic de-escalation. PUBLIC HEALTH RELEVANCE: Antibiotic de-escalation refers to the discontinuation of antibiotics that are providing a spectrum of bacterial activity greater than necessary to prevent the development of antibiotic resistance. This study seeks to develop a method for determining if antibiotic de-escalation has been practiced, measuring antibiotic de-escalation rates between hospitals, and determining factors that positively influence antibiotic de-escalation. The proposed research has significant potential to improve health by providing a basis for understanding relationships influencing antibiotic de- escalation practices.